Artificial intelligence (AI) is ‘progressively transforming’ the field of ophthalmic surgical training.
This is based on a recent review that evaluated artificial intelligence-assisted simulation and surgical video analytics for ophthalmic surgery training and competence development.
From experience-dependent apprenticeship to objective, data-driven competence development, AI plays a role.
And the review authors predict that AI-assisted ophthalmic surgical training will continue to shift from standalone tools to interconnected intelligent systems.
Current evidence supports the role of AI-enabled simulation, computer vision, and registry-driven predictive modelling as assistive tools that enhance feedback consistency, accelerate skill acquisition, and enable scalable benchmarking.
These systems work best when guided by clearly defined educational goals and human supervision, rather than acting as autonomous decision-makers, the authors write.
The next stage of AI development in ophthalmic surgery will move beyond task-specific models towards integrated, multimodal intelligence systems that function throughout the entire surgical process.
Future systems are expected to shift from single-modality video analysis to multimodal foundation models that jointly learn from surgical video, instrument kinematics, force-feedback signals, intraoperative OCT, and perioperative clinical data.
For training, this could allow a more detailed competency assessment that combines technical accuracy, timing, and mental effort. However, strong external validation and standardised data interfaces will be crucial before these models can be safely used at scale.
Another promising direction, the authors note, is the development of digital twins of ocular anatomy, created from patient-specific imaging and continuously updated surgical data.
Such virtual replicas could enable surgeons to practise procedures under realistic biomechanical conditions, explore alternative strategies, and anticipate potential complications before entering the operating theatre.
For trainees, digital twins may bridge the gap between generic simulation and personalised anatomy, speeding up the shift from rule-based execution to situational reasoning.
Instead of relying solely on expert behaviour, reinforcement learning-based systems could investigate action-outcome trade-offs in simulated environments to find safer or more efficient procedural pathways.
In the near term, such methods are best limited to offline simulation and training scenario design, where safety risks are minimal.
However, translating into real-world assistance will require strict safeguards, transparent reward systems, and clear boundaries that maintain surgeon authority.
Sustainable progress in AI-assisted surgical education will rely on access to diverse, high-quality data.
Federated learning provides a viable pathway to train robust models across institutions while preserving patient privacy and data sovereignty.
For ophthalmic surgery, federated infrastructures could minimise centre-specific bias, enhance generalisability across different devices and populations, and support international benchmarking of training outcomes.
More broadly, recent predictive modelling studies in other ophthalmic subspecialties have similarly highlighted the importance of diverse datasets and external validation, underlining the necessity for cross-centre collaboration before AI models can be reliably used as decision-support tools.
Beyond the scope of this review, AI-integrated robotic assistance and closed-loop control for ophthalmic microsurgery constitute a significant parallel pathway that will need dedicated evidence synthesis and prospective clinical validation.
The greatest impact will likely come not from full automation, but from thoughtfully designed human-AI collaboration that enhances perception, encourages reflection, and reinforces clinical responsibility, they conclude.
Future research should prioritise prospective, multi-centre studies that connect AI-enabled training interventions to long-term patient outcomes, thereby ensuring that technological progress results in meaningful clinical benefits.
You can read more here
